Claude Opus 5.5 vs Opus 5: What's New?
Everything that changed from Claude Opus 5 to Claude Opus 5.5: benchmarks, a 40% cost cut, faster output, clearer writing, and stronger safety.
The Verdict
"Upgrade. Opus 5.5 beats Opus 5 on every benchmark Anthropic published, while costing about 40% less and running 30% faster."
Anthropic released Claude Opus 5.5 on September 22, 2026, just two months after Claude Opus 5 (July 24, 2026). It's the first model in the Claude 5.5 family. It also runs on the same workloads for less money, less time, and fewer tokens.
Anthropic says Opus 5.5 performs at the level of its top-tier Claude Fable 5.1 on most work and costs 40% less to run than Opus 5.
At a Glance
| Claude Opus 5.5 | Claude Opus 5 | |
|---|---|---|
| Released | September 22, 2026 | July 24, 2026 |
| API price (input / output per M) | $4 / $20 | $5 / $25 |
| Cache reads (per M) | $0.20 | $0.50 |
| Typical cost per task | ~40% lower | baseline |
| Output speed | 30%+ faster | baseline |
| Context / max output | 1M / 128k | 1M / 128k |
| Knowledge cutoff | June 2026 | — |
| Default effort | medium | — |
| Terminal-Bench 4.0 | 66.4% | 52.3% |
| FrontierCode v1.1 | 54.4% | 48.0% |
| CursorBench 4.0 | 57.8% | 46.6% |
| GDPval-AA v2.1 | 1846 | 1708 |
| Humanity's Last Exam (with tools) | 67.7% | 63.6% |
| OSWorld 2.1 (computer use) | 81.8% | 74.0% |
Benchmarks from Anthropic's launch announcement. Opus 5.5 results use adaptive thinking at max effort unless noted.
What's New
1. Much Better at Agentic Coding
On Terminal-Bench 4.0, Opus 5.5 jumps 14 points over Opus 5. Anthropic reports that at default effort it already beats Opus 5 at max effort, for about a fifth of the cost.
"On our agentic coding tasks, Claude Opus 5.5 matched Opus 5's quality in about half the turns, time and output tokens, cutting the cost of that workload by 40 to 50%." — Optiver
2. Cheaper and Faster
List prices drop 20%, to $4 / $20. Cache reads drop 60%, to $0.20, which matters because they make up most of the cost of agentic and coding work. Opus 5.5 also uses fewer tokens per task, so typical workloads cost about 40% less overall, and output is over 30% faster.
Subscribers benefit too: Anthropic raised five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans.
3. Clearer Writing
The most common complaint about Opus 5 was how it wrote. Opus 5.5 puts the key information first, uses less jargon, and follows the writing rules you give it, which makes long sessions easier to follow.
4. Safer and Harder to Hijack
Opus 5.5 posts the best results Anthropic has published on its automated behavioral audit. It's much less likely to take hard-to-reverse actions and is more resistant than Opus 5 to prompt injection.
Should You Upgrade?
Yes, for almost everyone. Opus 5.5 is better on every benchmark Anthropic published and cheaper. In the Claude apps it's simply the Opus model you get.
API developers:
- Switch the model ID to
claude-opus-5-5. - The default effort is
medium. Benchmark there first before raising it, since it's often good enough and cheaper. - As with any model change, test your agent's prompts and tool interfaces on a sample of real traffic before switching production over.
Want to try it now? Start chatting with Claude Opus 5.5 with free messages and no sign-up needed.